State-Space Simulator: Phase Plane & Eigenvalue Analysis

simulator advanced ~12 min
Loading simulation...
λ = -1 ± j1.73 — stable spiral

With the default A matrix, eigenvalues are -1 ± j1.73, producing a stable spiral in the phase plane. States oscillate while decaying exponentially toward the origin.

Formula

ẋ = Ax + Bu (state equation)
λ = (tr(A) ± √(tr(A)² - 4·det(A))) / 2
Φ(t) = e^(At) (state transition matrix)

Beyond Transfer Functions

While transfer functions describe input-output relationships, state-space representations capture the complete internal dynamics of a system. By expressing an nth-order differential equation as n first-order equations, state-space reveals hidden modes, internal coupling, and structural properties like controllability and observability. This formulation, pioneered by Rudolf Kalman in the 1960s, became the foundation of modern control theory.

The Phase Plane

For a two-dimensional system, the phase plane plots x₁ against x₂, with each point representing a complete system state. Trajectories flow according to the vector field defined by ẋ = Ax, spiraling inward (stable focus), diverging outward (unstable focus), or following saddle-shaped paths. The eigenvalues of A completely determine the portrait: real eigenvalues give nodes, complex eigenvalues give spirals, and pure imaginary eigenvalues give concentric orbits (centers).

Eigenvalue Classification

The trace-determinant plane classifies all possible behaviors of a 2D linear system. Below the parabola tr²=4·det lie spiral behaviors (complex eigenvalues); above lie nodes (real eigenvalues). The tr=0 line separates stable (left) from unstable (right). The det=0 line marks the boundary where one eigenvalue is zero. This simulator visualizes the eigenvalues and maps them to the corresponding phase portrait in real time.

Control Design in State Space

State feedback u = -Kx places the closed-loop eigenvalues at arbitrary locations (if the system is controllable), a vastly more powerful technique than gain tuning alone. Combined with a state observer (Luenberger or Kalman filter), this yields the separation principle: design the controller and observer independently. These ideas underpin modern autopilots, robotic controllers, and industrial process control systems.

FAQ

What is state-space representation?

State-space represents a dynamic system as a set of first-order differential equations: ẋ = Ax + Bu, y = Cx + Du. The state vector x captures all information needed to predict future behavior. Unlike transfer functions, state-space handles multi-input multi-output (MIMO) systems and directly reveals internal dynamics.

What do eigenvalues tell us?

The eigenvalues of the system matrix A determine stability and dynamic character. Negative real parts mean stability; positive real parts mean instability. Complex eigenvalues indicate oscillation. The eigenvectors define the principal directions of motion in state space.

What is a phase plane portrait?

The phase plane plots one state variable against another, showing all possible trajectories from different initial conditions. Key features include equilibrium points, stable/unstable manifolds, and limit cycles. For 2D linear systems, the portrait is fully determined by the eigenvalues of A.

What is controllability?

A system is controllable if, for any initial and final state, there exists an input that drives the system between them. For the pair (A,B), controllability requires that the matrix [B, AB, A²B, ...] has full rank. Uncontrollable modes cannot be influenced by any input.

Sources

Embed

<iframe src="https://homo-deus.com/lab/control-systems/state-space/embed" width="100%" height="400" frameborder="0"></iframe>
View source on GitHub